Partnerships create the disaggregated ecosystem to support next generation digital transformation

Sponsored Article

The old saying is that it takes a village to raise a child. Well, it takes an ecosystem to enable a digitally transformed business. That is coming together now as hardware, software, apps, networking and vertical-specific skills create the businesses of the future. Sanjay Uppal, vice president and general manager of the software-defined edge business at VMware by Broadcom, tells IoT Now why one-size-doesn’t-fit-all anymore.

With Gartner predicting that 50% of enterprise data will be created outside a data centre by next year, it’s clear that next generation operations are happening at the edge and enabling the edge is now a priority for enterprises, he explains.

IoT Now: There are big changes happening across multiple industries as systems digitise and enterprises look to align their infrastructure with their new needs. How complex is this transformation and what should organisations be looking for in a partner?

Sanjay Uppal: It depends on the industry you’re in. Some are already transformed while others are just getting started. For example, the telecoms industry first started its digital transformation route by disaggregating the core and has moved on to the radio access network (RAN). The benefit is agility in rolling out new apps – but to get that disaggregation done you need platform partners with two key skills.

First, they must be a great ecosystem player that is able to bring together apps, hardware and services. Second, they must be able to provide digital infrastructure that scales. Providing the ecosystem and infrastructure that scales are the two hallmarks of an effective partner, but in industrial manufacturing, healthcare, logistics and many other verticals.

IoT Now: To take one example of the transformation burden, how do you see organisations’ efforts to right-size their infrastructure in the face of moves to software-defined edge and away from monolithic IT architectures? There are several different dynamics involved in this but each is interrelated, so how can holistic approaches that keep projects on-track be enabled?

SU: You start off with where the devices that are producing or consuming data are – and more than half will be outside the data centre. You’ll find that most of these workloads outside the datacentre are highly specialised and constrained by hardware. After you locate your use cases, you then look at the functionalities you need to place close to those devices to enable them.

Why does being close matter? Because latency is the new currency whether for self-driving cars, robots on the factory floor or for customer loyalty in a retail shop.

You start off with a blueprint for digital transformation and then build the ecosystem you need to enable the use cases you’ve identified. For a retail store, you can chart a path from being inefficient to automatically replenishing shelves when stock goes down and reducing shrinkage by identifying live theft.

In grocery chains in the UK, for example, people can walk in and out without engaging with a sales representative or scanning goods. Through a combination of edge devices, the store recognises the items they’ve shopped for and charges them for it. That’s a computer vision use case and to achieve that you need specific partners.

Regardless of the use case, the process begins in the same way – identifying what the sources and consumers of data are and then rolling-out the ecosystem based on those visions.

IoT Now: The nature of IoT means there is a ‘things’ part of the transformation so the work can’t just be done in the internet or the data centre. What challenges does having to upgrade infrastructure on the factory floor involve?

SU: There are several challenges to consider. Location and proximity to similar workloads is one, latency is another and you can’t have an intelligent conversation about the edge without talking about the network underneath. Today, there are industry sites that have devices numbering in the millions that require a level of scale that is very different from a data centre.

The first challenge is to right-size the infrastructure so it can run on very small hardware and we’ve been able to do that with the VMware Edge Compute Stack. It really is a case of ‘Honey, I shrunk the stack’.

For example, with new machine learning workloads, it’s possible to run models on hardware that has less than 256k of RAM. In this case you could have a tiny sensor on the side of a machine that measures a rise in temperature or vibration and can predict when the machine will fail. That’s a very different scenario than a large language model (LLM) but still has a high impact since you can act on the insight very quickly and avert a catastrophe in the next few minutes.

On a factory floor there are not enough IT staff to understand if a potential interference has been added to the floor, but a simple sensor can provide valuable insight to defend against operational challenges. A sensor doesn’t need more computing power or a high-capacity connection. It just needs to be able to react to performance data outside of a preset range and communicate that. Moreover, these kinds of workloads can be consolidated on edge devices to further shrink the stack. While we aren’t talking about the thousands of workloads in the data centre, we do see consolidation of about 6-to-1.

One thing that is critically important from the edge standpoint is that all things behave as if they are independent. It’s therefore essential to come up with a way to synchronise management of edge devices and intelligence. We’ve developed our VMware Edge Cloud Orchestrator to achieve this so when the edge device wakes up it can call home and get access to the platform software as needed.

Being able to shrink the footprint and ensure security are all hallmarks of being able to run efficiently in very diverse environments which bring new requirements. For example, in manufacturing, safety and compliance is very important. The machine should be shut off immediately if it is failing whereas in a retail site, there are no moving parts so there’s no need for millisecond-level responses.

The technology that supports this is made up of the vertically-oriented ecosystem that provides specific software, hardware and services for that use case, such as app providers or computer vision specialists, and vendors like us who have the ability to put the entire ecosystem together on a vertical by vertical basis along with the telecoms operators, who are essential to edge performance.

IoT Now: How do you see the software-defined wide area network (SD-WAN) providing the connectivity tissue that complex operational environments, such as factories, need? How challenging is it to encapsulate the diverse connectivity needs of different equipment in a modern, fit for purpose infrastructure?

SU: SD-WAN started off being the solution for the connectivity for enterprise branch offices to the data centre. The solution we built excelled at that but now the effort is focused on how SD-WAN can be the fabric for the next generation workloads of digital transformation. That’s comprised of low latency, orchestration and the best quality-of-service for the apps you want. These are the challenges that SD-WAN has to extend into.

I view it as the next era of SD-WAN as it’s all about programmability because the needs of office apps or Office365 are very different to factory or hospital use cases. These new apps see SD-WAN as the way to bring all the different requirements together. Systems don’t have to tell if a device is a screwdriver on a factory floor, for example. The device is already recognised but the network has to be programmed to provide the right QoS for the use case.

This is the advantage for telecoms operators and providers of the underlying technology because they can add value. In addition, with technologies such as low earth orbit (LEO) satellites becoming available there is another link available to support all these new capabilities. Regardless of this, you need an intelligent layer so a device can say: ‘I’m a respirator and I need a response within one second’ or ‘I’m a sensor uploading non-urgent shift-end data.’

I’m very excited about this next phase with SD-WAN.

IoT Now: There’s understandable caution regarding investing in the next generation connectivity infrastructure to support IoT-enabled operations. How does Broadcom give customers confidence that they will not only achieve greater agility and flexibility today with their transformations but also be well-placed for these networks to support their future needs?

SU: The glib answer in computer science is to add another layer of abstraction. Here we have another layer of programmable software and also the edge compute stack. These layers allow innovation to happen much faster than if you deployed monolithic systems. There is a reason architectures are moving towards disaggregation and software- defined – to support unanticipated growth in the future while achieving agility and flexibility today. Broadcom’s tagline is ‘connect everything’ and what we are doing at the software-defined edge fits in with that objective. Other divisions are working on next gen GPUs and producing encoders for next generation networks. We’re connecting everything, and that’s what is required to thrive in the future.

IoT Now: Insofar as it’s possible to define specific attributes given the diversity of transformations, what’s the right partner attitude here?

SU: Usually when a new set of technologies comes in you get companies trying to do everything from soup to nuts, but I don’t think that works here, especially with AI and machine learning. There are a wide variety of use cases that showcase an enormous array of diversity. That diversity means there is not going to be a one-size-fits-all offering so an ecosystem approach is needed. You need a partner that is building the ecosystem with hardware partners, telcos, app providers, hyperscalers with next generation programmability, and specialists in small language models.

That sounds like a lot of partners, but consider the use cases and what drives selections of a particular blend of partners. For instance, a hardware provider that allows us to apply the software stack or a systems integrator that knows the business and can simplify use case projects. There will be many players at the edge, so finding partners who enable this ecosystem-approach is important to success.

IoT Now: What is Broadcom’s strategy for helping customers transform and optimise their operations for a new era?

SU: At Broadcom, we understand where we can add value and that’s not by providing vanilla offerings. We’re a world class technology provider and that starts off by understanding what the customer’s use case is trying to achieve.

Our approach to all of our engagements is to achieve simplification and disaggregation, and to abstract complexity away. This is how we run our business internally and for our customers. Our software-defined edge portfolio is a great example of this. You bring the capabilities to where you need them, when you need them, so your business can be more efficient and agile. We’re doing this so customers can concentrate on their businesses rather than their technology.

Comment on this article via X: @IoTNow_ and visit our homepage IoT Now